Page 1 of 1

Failover and Performance Enhancement

Posted: Mon Sep 09, 2013 7:13 am
by Upatel
Could you clarify a few things for me in reference to this diagram:

Is the "MOD-GEARMAN WORKER" box to the right, just and example of an extra worker, or is it a requirement to implement MOD-GEARMAN?
Can the "PUPPET MASTER" server be used to manage configurations on both Production and Test environments, or would you recommend a separate one for Production?
If you have moved the MySQL databases and Postgres database to a separate server (not on the Nagios XI Server), would this configuration below still work or would there be modifications?

Re: Failover and Performance Enhancement

Posted: Mon Sep 09, 2013 9:44 am
by abrist
Upatel wrote: Is the "MOD-GEARMAN WORKER" box to the right, just and example of an extra worker, or is it a requirement to implement MOD-GEARMAN?
It is under the "performance" section and is not required.
Upatel wrote:Can the "PUPPET MASTER" server be used to manage configurations on both Production and Test environments, or would you recommend a separate one for Production?
It can be used for both, you just have to make sure to configure puppet to do so.
Upatel wrote:If you have moved the MySQL databases and Postgres database to a separate server (not on the Nagios XI Server), would this configuration below still work or would there be modifications?
This configuration should work, at least for an offloaded mysql database. postgres is not so easily offloaded and shared though.

Re: Failover and Performance Enhancement

Posted: Sun Sep 15, 2013 11:31 am
by mikew
One of the keys to this design is that Mod-Gearman is running on both the Master and the Slave. The Slave has two queues one for the Master which will have checks while the Master is running. The second queue is for the Slave which will not have any jobs while the Master is running. When the Master stops, the jobs in the Master queue will stop, and the Slave takes over and the jobs will be for the Slave in that queue. So Mod-Gearman is working on both Master/Slave and will help distribute load regardless of which is running. This also means that if you have a failover server it is actually doing something when things are good...justifies the cost.

Re: Failover and Performance Enhancement

Posted: Mon Sep 16, 2013 10:27 am
by lmiltchev
Upatel, do you have any more questions?